Models Manager

The Models Manager is where you can manage and deploy published machine learning models.

Models Manager


Getting There

Access Arundo Fabric and select Models on the Navigation bar.


Resources Panel

The Resources panel displays real-time metrics about system resource usage by published models.

Warning

If resource usage is high, you may be unable to publish new models and could experience performance degradation with existing models.

Resources panel


Models List

The Models list displays details for the models currently published to Arundo Fabric.

Important

You must publish a model into Arundo Fabric from Arundo Composer before it displays in this list.

Models list

The total number of models currently published displays in the top right corner of the list.

Fields

The following information displays for each model:

Field Description
Filter by owner Filters the list to show the models published by the specified user
Name Name of the model
Version Current version deployed
Owner User who published the model
Created Time elapsed since the model was published
Status Indicates whether the model is currently published to Arundo Fabric
Actions

The following actions are available for each model:

Action Description
Delete Remove the model from Arundo Fabric
Share Copy the URL for the model application so you can share the link with other users
Deploy to Edge Deploy the model to Arundo Edge
Use Open the model application

Model Details

The Model Details page displays when you click a Name in the Models list.

Models Details page

Use the Model Details page to view detailed specifications and diagnostic information for a model.

Model Details

The Model Details panel displays the following information about the model:

Field Description
Model Name Name of the model
Version Current version of the model published
Status Indicates whether the model is currently published to Arundo Fabric
Created Time elapsed since the model was published
Owner User who published the model
ID Unique identifier for the model
Scale Indicates how many identical copies of the model are currently published to Arundo Fabric
CPU Current CPU limit for Arundo Fabric
Memory Current memory request limit for Arundo Fabric
Version Control

The Version Control panel displays GitHub commit details for the model, if it was published from a GitHub repo.

Field Description
Remote Name of the model
Hash Commit hash for the merge that published the model
Author User who published the model
Environment Variables

The Environment Variables panel give you the ability to add or edit environment variables for the model.

Tip

An environment variable gives you the ability to change the database connection string for a model without having to redeploy it.

Model Log

The Model Logs panel displays an activity log for the model. This is useful when troubleshooting publishing issues.


Model Application

A model application is a web browser-based app you can use to interact with a model. This lets you test a model locally in a controlled environment.

Home page

Getting There

  1. Open the Models Manager.
  2. Locate the model in the Models list.
  3. Click Use.
  4. Enter your Arundo Fabric login credentials.

Index

The Index page displays when you first open a model application. From here, you can test the model and view its inputs and outputs.

Index page

Tips

  • Click Use... to enter inputs and calculate the model's output.
  • Click Describe to display the inputs and outputs for the model.

Developer

The Developer page is where you can view the model's inputs and outputs, and test the model.

Developer page

Tips

  • Click the URL for the GET method to display the inputs and outputs for the model.
  • Click the URL for the POST method to enter inputs and calculate the model's output.

Endpoints

The Endpoints page is where you can view the model's inputs and outputs.

Endpoints page

Tip

Click Expand to display inputs and outputs.


Step-by-Step Guides

The following guides explain how to test and deploy models.

Testing a Model

Complete the following to test a model in a model application.

Prerequisites

  • Publish the model into Arundo Fabric from Arundo Composer

Instructions

  1. Locate the model in the Models list.
  2. Click Use next to the model.
  3. Enter your Arundo Fabric login credentials.
  4. Click Login.
  5. Click Use....
  6. Enter input values.
  7. Click Execute Model.
Deploying a Model to Arundo Edge

Complete the following to deploy a machine learning model to Arundo Edge. This gives you the ability to run tag data through the model and view visualizations of the model output in Arundo Edge.

Prerequisites

  • Verify your device meets the system requirements​
  • Install the latest version of Docker Desktop/CE (Windows / macOS / Linux)
  • Start Docker and keep it running
  • Publish the model into Arundo Fabric
  • Set up a Cloud interface between Arundo Edge and Arundo Fabric

Instructions

  1. Access Arundo Edge.
  2. Open the System menu and select Upload Configuration to Fabric.

    Upload Configuration

  3. Access Arundo Fabric.

  4. Select Models on the Navigation bar.
  5. Locate the model in the Models list.
  6. Click Deploy to Edge next to the model.
  7. In the Edge Agents list, select the edge device where you want to deploy the model.
  8. Under Endpoint Configuration, select the tags to use for the model's inputs and outputs.
  9. Click Deploy.

    Tip

    To view a visualization of the model's output in Arundo Edge, go the Viewer page and look under Model.